Derive period-scoped treatments and dates across a washout

1library(yamaa)23yamaa_spec("spec.yaml")$run()$output

1import yamaa23yamaa.spec("spec.yaml").run().output

1# Reference solution for the yamaa benchmark adam-adsl-crossover-periods (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)910dm <- read_csv(11  "/app/input/dm.csv",12  col_types = cols(.default = col_character())13)14ex <- read_csv(15  "/app/input/ex.csv",16  col_types = cols(17    EXSEQ = col_integer(),18    EXSTDTC = col_date(),19    EXENDTC = col_date(),20    .default = col_character()21  )22)2324ex1 <- ex |> filter(EPOCH %in% "TREATMENT 1")25ex2 <- ex |> filter(EPOCH %in% "TREATMENT 2")2627agg1 <- ex1 |>28  group_by(STUDYID, USUBJID) |>29  summarise(30    TR01SDT = if (all(is.na(EXSTDTC))) as.Date(NA) else min(EXSTDTC, na.rm = TRUE),31    TR01EDT = if (all(is.na(EXENDTC))) as.Date(NA) else max(EXENDTC, na.rm = TRUE),32    .groups = "drop"33  )34agg2 <- ex2 |>35  group_by(STUDYID, USUBJID) |>36  summarise(37    TR02SDT = if (all(is.na(EXSTDTC))) as.Date(NA) else min(EXSTDTC, na.rm = TRUE),38    TR02EDT = if (all(is.na(EXENDTC))) as.Date(NA) else max(EXENDTC, na.rm = TRUE),39    .groups = "drop"40  )4142# The treatment on the earliest record in each period: earliest start date43# with the lower sequence number breaking ties, records without a start44# date sorting last.45first1 <- ex1 |>46  arrange(STUDYID, USUBJID, is.na(EXSTDTC), EXSTDTC, EXSEQ) |>47  distinct(STUDYID, USUBJID, .keep_all = TRUE) |>48  transmute(STUDYID, USUBJID, TRT01A = EXTRT)49first2 <- ex2 |>50  arrange(STUDYID, USUBJID, is.na(EXSTDTC), EXSTDTC, EXSEQ) |>51  distinct(STUDYID, USUBJID, .keep_all = TRUE) |>52  transmute(STUDYID, USUBJID, TRT02A = EXTRT)5354adsl <- dm |>55  left_join(agg1, by = c("STUDYID", "USUBJID")) |>56  left_join(agg2, by = c("STUDYID", "USUBJID")) |>57  left_join(first1, by = c("STUDYID", "USUBJID")) |>58  left_join(first2, by = c("STUDYID", "USUBJID")) |>59  mutate(WASHDUR = as.integer(TR02SDT - TR01EDT) - 1L) |>60  select(61    STUDYID, USUBJID, TR01SDT, TR01EDT, TR02SDT, TR02EDT,62    TRT01A, TRT02A, WASHDUR63  )6465dir.create("/app/output", showWarnings = FALSE)66write_csv(adsl, "/app/output/adsl.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adsl-crossover-periods (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67"""Build ADSL (one record per subject) from DM and EX.89Derives the two crossover treatment periods per subject: earliest start10and latest end per epoch, the treatment of the earliest record in each11epoch, and the washout duration strictly between the periods.12"""1314import csv15from datetime import date1617INPUT_DIR = "/app/input"18OUTPUT_DIR = "/app/output"1920COLUMNS = [21    "STUDYID", "USUBJID", "TR01SDT", "TR01EDT", "TR02SDT", "TR02EDT",22    "TRT01A", "TRT02A", "WASHDUR",23]242526def parse_day(text):27    text = (text or "").strip()28    if not text:29        return None30    return date.fromisoformat(text)313233def read_rows(path):34    with open(path, newline="") as handle:35        return list(csv.DictReader(handle))363738def period(rows):39    """Return (start, end, treatment) for one subject's records in one epoch."""40    starts = [d for d in (parse_day(r["EXSTDTC"]) for r in rows) if d]41    ends = [d for d in (parse_day(r["EXENDTC"]) for r in rows) if d]42    start = min(starts) if starts else None43    end = max(ends) if ends else None44    treatment = ""45    if rows:46        def earliest_key(row):47            day = parse_day(row["EXSTDTC"])48            # A record with no start date counts as later than any with one.49            return (day is None, day or date.max, float(row["EXSEQ"]))5051        treatment = min(rows, key=earliest_key)["EXTRT"]52    return start, end, treatment535455def main():56    dm = read_rows(f"{INPUT_DIR}/dm.csv")57    ex = read_rows(f"{INPUT_DIR}/ex.csv")5859    by_subject = {}60    for row in ex:61        by_subject.setdefault((row["STUDYID"], row["USUBJID"]), []).append(row)6263    out = []64    for subject in dm:65        key = (subject["STUDYID"], subject["USUBJID"])66        records = by_subject.get(key, [])67        t1 = [r for r in records if r["EPOCH"] == "TREATMENT 1"]68        t2 = [r for r in records if r["EPOCH"] == "TREATMENT 2"]69        tr01sdt, tr01edt, trt01a = period(t1)70        tr02sdt, tr02edt, trt02a = period(t2)7172        washdur = ""73        if tr01edt is not None and tr02sdt is not None:74            # Days strictly between the two dates, counting neither endpoint.75            washdur = str((tr02sdt - tr01edt).days - 1)7677        def fmt(day):78            return day.isoformat() if day is not None else ""7980        out.append({81            "STUDYID": subject["STUDYID"],82            "USUBJID": subject["USUBJID"],83            "TR01SDT": fmt(tr01sdt),84            "TR01EDT": fmt(tr01edt),85            "TR02SDT": fmt(tr02sdt),86            "TR02EDT": fmt(tr02edt),87            "TRT01A": trt01a,88            "TRT02A": trt02a,89            "WASHDUR": washdur,90        })9192    with open(f"{OUTPUT_DIR}/adsl.csv", "w", newline="") as handle:93        writer = csv.DictWriter(handle, fieldnames=COLUMNS)94        writer.writeheader()95        writer.writerows(out)969798if __name__ == "__main__":99    main()

1/* opensas reference solution for the yamaa benchmark adam-adsl-crossover-periods.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data dm;6  length STUDYID USUBJID $1024;7  infile "/app/input/dm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024.;9run;1011data ex;12  length STUDYID USUBJID EPOCH EXTRT EXSTDTC EXENDTC $1024;13  infile "/app/input/ex.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input STUDYID : $char1024. USUBJID : $char1024. EXSEQ EPOCH : $char1024. EXTRT : $char1024. EXSTDTC : $char1024. EXENDTC : $char1024.;15run;1617data exposure; set ex; NODATE=missing(EXSTDTC); run;18proc sort data=exposure; by STUDYID USUBJID EPOCH NODATE EXSTDTC EXSEQ; run;19data firstex; set exposure; by STUDYID USUBJID EPOCH; if first.EPOCH; run;20proc sql;21create table starts1 as select STUDYID,USUBJID,min(EXSTDTC) as TR01SDT from ex where EPOCH='TREATMENT 1' and not missing(EXSTDTC) group by STUDYID,USUBJID;22create table ends1 as select STUDYID,USUBJID,max(EXENDTC) as TR01EDT from ex where EPOCH='TREATMENT 1' group by STUDYID,USUBJID;23create table starts2 as select STUDYID,USUBJID,min(EXSTDTC) as TR02SDT from ex where EPOCH='TREATMENT 2' and not missing(EXSTDTC) group by STUDYID,USUBJID;24create table ends2 as select STUDYID,USUBJID,max(EXENDTC) as TR02EDT from ex where EPOCH='TREATMENT 2' group by STUDYID,USUBJID;25create table joined as select a.*,b.TR01SDT,c.TR01EDT,d.TR02SDT,e.TR02EDT,f.EXTRT as TRT01A,g.EXTRT as TRT02A from dm a26left join starts1 b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID27left join ends1 c on a.STUDYID=c.STUDYID and a.USUBJID=c.USUBJID28left join starts2 d on a.STUDYID=d.STUDYID and a.USUBJID=d.USUBJID29left join ends2 e on a.STUDYID=e.STUDYID and a.USUBJID=e.USUBJID30left join firstex f on a.STUDYID=f.STUDYID and a.USUBJID=f.USUBJID and f.EPOCH='TREATMENT 1'31left join firstex g on a.STUDYID=g.STUDYID and a.USUBJID=g.USUBJID and g.EPOCH='TREATMENT 2'; quit;32data result; set joined; WASHDUR=.; if not missing(TR01EDT) and not missing(TR02SDT) then WASHDUR=max(0,input(TR02SDT,yymmdd10.)-input(TR01EDT,yymmdd10.)-1); run;33proc export data=result(keep=STUDYID USUBJID TR01SDT TR01EDT TR02SDT TR02EDT TRT01A TRT02A WASHDUR) outfile="/app/output/adsl.csv" dbms=csv replace; run;